CN104036499A - Multi-scale superposition segmentation method - Google Patents
Multi-scale superposition segmentation method Download PDFInfo
- Publication number
- CN104036499A CN104036499A CN201410238491.3A CN201410238491A CN104036499A CN 104036499 A CN104036499 A CN 104036499A CN 201410238491 A CN201410238491 A CN 201410238491A CN 104036499 A CN104036499 A CN 104036499A
- Authority
- CN
- China
- Prior art keywords
- yardstick
- scale
- segmentation
- characteristic parameter
- data layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a multi-scale superposition segmentation method, which comprises the following steps that: images are subjected to multi-scale segmentation to obtain each scale object after the segmentation; a stable scale index is utilized for judging each scale object to determine the optimum scale object of each scale; the determined optimum scale object of each scale is projected to a single data layer; and the optimum scale objects projected onto the single data layer are merged. The multi-scale superposition segmentation method has the advantages that the segmentation object extraction is carried out by using real ground object matching based on each object boundary as targets; the multi-scale segmentation on the images is realized; the problems of over-segmentation and under-segmentation of objects in the same scale but different land coverage types and the problem of overlapping during the multi-scale object superposition are avoided; and the precision of the multi-scale segmentation is improved.
Description
Technical field
The present invention relates to the Satellite Remote Sensing field in geography, relate in particular to a kind of multiple dimensioned stack dividing method.
Background technology
The spectral characteristic of remote sensing image greatly affects Land Cover Mapping precision.This has similarity between inside heterogeneity highly and class different soils cover type in spectral signature, even high resolution image, also often owing to producing " same object different images " phenomenon, reduce nicety of grading, for traditional sorting technique based on pixel, single yardstick spectrum, be difficult to solve the problem of nicety of grading.
Scale effect refers to the information translation in different time and space scales or different tissues level, the theoretical and ground process angle based on metric space, and on different scale, often there is different characteristic rules in pattern and process.Be subject to the impact of sensor imaging pattern, the space scale feature of remote sensing image and land cover classification system organization scale feature (pressing the system that pedigree structure is set up) are often not exclusively identical, and in the same grade of categorizing system, each soil cover type can not effectively characterize on the same space image yardstick.And utilize same yardstick remote sensing image to carry out soil and cover monitoring, can cause the inconsistent of each soil cover type nicety of grading.Different soils cover type is different to the basis of space scale, stability.All types of have different best observed range and a yardstick, could be effectively, intactly observe, and might not be that distance is more closely better, observation is more trickle better, the space scale of single optimization is difficult to the soil cover type under the complicated image of accurate Characterization.
In choice of optimal scale and sort research, proposed the earliest to determine image optimum segmentation yardstick with imaged object mean variance method.By forming the variance of whole region/image that average produces of all pixel brightness values of this object, set up the curve of multiple dimensioned variance, determine that its peak value has extracted different classes of its corresponding optimum segmentation yardstick that has, the method is particularly useful for the selection of high resolution image optimal scale.Afterwards, propose again to determine image optimum segmentation yardstick with imaged object maximum area method.Imaged object maximum area is with the trend of cutting apart the stepped rising of curve of dimensional variation, and each curve platform is corresponding to the scope of the suitable yardstick of certain classification extraction.Wherein, when analyzing Lidar altitude information and spectrum, texture, shade dimension relation property, equidistantly cut apart 15 yardsticks, analyze the correlativity under different scale, the best scale dimension applications of correlativity is best scale.Due to dissimilar, on different scale, one of them type is at best yardstick, and other type may be in over-segmentation and less divided phenomenon, so for single scale classification, select best scale to be actually the average fitness scale of selecting most of types.
To this, improved method is the dissimilar method of classifying of matching on different scale, i.e. multiple dimensioned matching classification.As utilize the method for visual analysis of experiments, select Dan Shu, woods spot, 3 yardsticks of Landscape Characteristics to extract different characteristic, same Forest Types.Or on 3 yardsticks, utilize SVM (Support Vector Machine, support vector machine) method to extract respectively the 3 class city impermeable surfaces such as wide highway, path, buildings.Because multiple dimensioned selection is with subjectivity, randomness, nonrepeatability, while analyzing for single class, be feasible, and difficulty is larger while analyzing for most classifications.Be that these class methods do not solve the overlap problem occurring when multiple dimensioned classification results superposes, lap is the result of high precision yardstick preferentially.
At present, utilize research that multiple dimensioned object-oriented method carries out land cover classification in the stage at the early-stage, the scaling research that soil is covered to imaged object feature is still faced with following problem: (1) is because spectrum and the space characteristics of soil cover type are heterogeneous, and the regional differentiation of same type, rule and the mechanism of the dimensional variation that different soils cover are also unclear, also do not form the scale selection method of a kind of robustness, standard; (2) dimensional variation feature is not also fully excavated, and classification mainly depends on two-dimentional spectrum and several how feature, has ignored in scaling process characteristic use longitudinally; (3) multiple dimensioned land cover classification do not form association, during classification results overlapping caused separately final synthetic classification results, produces propagation of error, and the multiple dimensioned soil of multiclass coverage effect is unsatisfactory.
Summary of the invention
The object of the embodiment of the present invention is to provide a kind of multiple dimensioned stack dividing method, by the true units match based on each object bounds, be that target is carried out cutting object extraction, realized the multi-scale division to image, the appearance of the overlap problem while having avoided the stack of multiple dimensioned object, has improved the accuracy of multi-scale division.
In order to achieve the above object, the embodiment of the present invention provides a kind of multiple dimensioned stack dividing method, said method comprising the steps of:
Image is carried out to multi-scale division, each yardstick object after being cut apart;
Utilization is stablized yardstick index and is judged each yardstick object, determines the best scale object of each yardstick;
The best scale object of each yardstick of determining is projected to single data Layer;
The best scale object projecting on single data Layer is merged.
Preferably, described image is carried out to multi-scale division, specifically comprise: based on region Fusion, from pixel, become object, small object to be fused into large object pixel fusion, merge step by step.
Preferably, after each yardstick object that obtains cutting apart, in described each yardstick object, include respectively characteristic parameter separately, wherein, object standard deviation SD is the parameter of best scale identification.
Preferably, what take over-segmentation object cuts apart yardstick at first as border, and co-located is cut the object of each yardstick, keeps each yardstick object characteristic parameter separately, so that different scale object carries out the contrast of co-located characteristic parameter.
Preferably, the characteristic parameter of each yardstick object of yardstick index analysis is stablized in described utilization, specifically by following formula:
S
i＝F
i+1-F
i
Wherein, S
irefer to yardstick index of stability, i refers to level of zoom, F
ithe SD value of object on i yardstick, F
i+1to cut apart the object SD value in rank at i+1 yardstick; In whole dimensional variation, S
ibe 0 continuously and continue when the longest, this interval scale is defined as best object fitting yardstick.
Preferably, described the best scale object projecting on single data Layer is merged, specifically comprise: the characteristic parameter of single yardstick object of take is to merge according to the same value of carrying out adjacent object.
Prior art is compared, and technical scheme that the embodiment of the present invention proposes has the following advantages:
The above embodiment of the present invention, by the multi-scale division of image, true units match based on each object bounds is that target is carried out cutting object extraction, avoided on same yardstick, over-segmentation and the less divided of different soils cover type object, the appearance of the overlap problem during stack of multiple dimensioned object, has improved the accuracy of multi-scale division.
Accompanying drawing explanation
Fig. 1 is the schematic diagram of the existing middle multi-scale division process that provides of the embodiment of the present invention;
Fig. 2 is the schematic flow sheet that multiple dimensioned stack that the embodiment of the present invention provides is cut apart;
Fig. 3 is the pictorial diagram that multiple dimensioned stack that the embodiment of the present invention provides is cut apart;
Fig. 4 is the segmentation effect figure that multiple dimensioned stack that the embodiment of the present invention provides is cut apart.
Embodiment
Below in conjunction with the accompanying drawing in the present invention, the technical scheme in the present invention is clearly and completely described, obviously, described embodiment is only a part of embodiment of the present invention, rather than whole embodiment.Embodiment based in the present invention, those of ordinary skills, not making all other embodiment that obtain under creative work prerequisite, belong to the scope of protection of the invention.
In existing multi-scale division, be the process of object (vector) that pixel (grid) is become, its object is, when type is divided, not only to consider the spectral signature of target, it is also conceivable that the feature such as shape, spatial relationship of target that object produces, thereby improved nicety of grading.Referring to lower Fig. 1, it is the schematic diagram of this multi-scale division process; Multi-scale division process is by the setting of a yardstick threshold value, obtains the mode of multistage pixel fusion.Concrete, cutting apart is to start to merge from pixel, and along with yardstick constantly increases, object constantly increases, and it is from partial pixel merging, cover type unit, whole soil, to the process of the combination of a plurality of unit.In different phase, object components is different, thereby show different characteristics of objects, and best yardstick is exactly, object size (primitive) is with truly border, atural object unit (target) is consistent, object spectrum now, how much, be related to that semantic feature is the feature that truly reflects atural object, utilize this feature to classify, be conducive to improve the nicety of grading of image.
The present invention is the basic ideas of the optimum segmentation scale selection based on single object: when equidistant yardstick threshold value increases, pixel constantly merges to large object to object, object, although not necessarily each changes of threshold all can have the change of object size, until there is a yardstick (threshold value), size and the real goal of object match, within the scope of some scale, it is stable or constant that object size can keep.Difference between two types is larger, and stable range scale is wider.When cutting apart yardstick, continue to increase, than real-world objects, large and two soils cover class objects and merge object size, and characteristics of objects also constantly changes thereupon.In dimensional variation, there are a plurality of stable yardsticks, and maximum (the widest) yardstick wherein, can be considered the best scale of target matching, also be this unit optimal segmentation state, by the best object extraction on these different scales out, project in a data plane, form optimal segmentation object layer, be conducive to further land cover classification of later stage.
Referring to Fig. 2, the schematic flow sheet that the multiple dimensioned stack providing for the embodiment of the present invention is cut apart, Fig. 3 is specially the figure signal obtaining according to the process flow diagram of Fig. 2 multi-scale division.
This flow process can comprise:
Step 201, carries out multi-scale division to image, each yardstick object after being cut apart.
In this step, image is carried out to multi-scale division, specifically comprise: based on region Fusion, from pixel, become object, small object to be fused into large object pixel fusion, merge step by step.
After each yardstick object after being cut apart, also comprise: the object bounds according to described over-segmentation yardstick cuts each yardstick object, each yardstick object after being cut, the object after cutting keeps original SD feature.
Concrete, utilize region folding, carry out the multiple dimensioned cutting procedure from bottom to top, from a pixel to object, this process can realize by Definiens software.In iterative step repeatedly, the control based on nonuniformity threshold value, less image object is merged into larger object, and multi-scale division is followed pedigree process, and the Edge keeping of over-segmentation yardstick is in less divided yardstick border.Change threshold value and represent to change scale size.Multi-scale division, since 0, is carried out yardstick lifting with the equidistant threshold value of benchmark (being generally " 5 ") increment.Parameter in " 5 " yardstick threshold range changes little, and is the yardstick tolerance that enough narrow scope is carried out Measurement sensibility.From the software of Definiens, export object layer each yardstick, that comprise SD information, and import ARCGIS software (a kind of vector space analysis software) and carry out the contrast in the space of multiple dimensioned object.All yardstick data Layers cut with over-segmentation (the closeest dividing layer) benchmark.Thereby the object boundary line that forms unified over-segmentation yardstick layer, is assigned to the SD feature of former object in the object properties of new cutting.This process guarantees that the projection of follow-up object does not produce overlapping and empty between object.
Step 202, utilization is stablized yardstick index and is judged each yardstick object, determines the best scale object of each yardstick.
In this step, in described each yardstick object, include respectively characteristic parameter separately, described utilization is stablized yardstick index and is judged each yardstick object, determines the best scale object of each yardstick.
Concrete, the characteristic parameter comprising in each yardstick object is specially standard deviation SD, pixel average etc.In this step, specifically usining characteristic parameter sets forth as preferred embodiment as SD.
Further, the standard deviation SD of each object is specially the spectral value statistics of each pixel in object, it stablizes the signature analysis of yardstick as eigenwert, consider that image is comprised of a plurality of wave bands, in fact the SD of object refers to the Euclidean distance (SD of the object of each wave band square root sum square) of all image spectral bands, selects SD more responsive than the size (weak relevant) of object average (normally fluctuation changes) or object as scale parameter (SD increases along with yardstick increases conventionally).Subsequently, SD multiple dimensioned in analytic target attribute list changes, and extracts best scale.
From a yardstick to the variation of another yardstick, assess SD and change, can utilize yardstick index of stability (Si) to express:
S
i＝F
i+1-F
i (1)
Wherein, S
irefer to yardstick index of stability, i refers to level of zoom, F
ithe level that the SD value of object is cut apart in expansion, F
i+1be that the yardstick of the i+1 of object is cut apart rank.
The attribute of the SD of each layer of object is input to MATLAB software (numerical value process software) by dimensional variation order.Calculate each adjacent yardstick S
i(formula 1), S
ivalue equals zero or is zero to occur the stable yardstick representing continuously, is wherein in the yardstick of zero width maximum continuously, automatically selects this width section medium scale to be expressed as best scale.
Step 203, projects to single data Layer by the best scale object of each yardstick of determining.
Concrete, above-mentioned definite best scale being identified, and SD corresponding to these best scale extracted, assignment is in an independent data Layer.
Step 204, merges the best scale object projecting on single data Layer.
Concrete, the SD of take carries out the similar merging of spatial neighbor object bounds as attribute, and these vector borders after merging are exactly optimal segmentation border, and these partitioning boundaries are finally input in Definiens software, image is cut apart, then extracted the work and rest of image spectrum and carry out Images Classification.
After carrying out multi-scale division, in order to verify segmentation effect, choose all kinds of soils cover type below and carry out recruitment evaluation.Select 8 common class soil cover types, the arable land, fallow ground, the water surface, residence, the traffic land used that comprise coniferous forest, broad-leaf forest, meadow, plant growth, from 5 scale parameters (over-segmentation yardstick) image, 10 objects of every class random acquisition, totally 80 objects, on this basis, 24 yardsticks of multi-scale division.In the analytical approach of real best scale service test-error, select best yardstick, target actual size and imaged object edge fitting are best scale.Object SD is along with yardstick increases, and increase or constant gradually.There is the SD change of three types: unstable (variation continuously, S
i>0); Relatively stablely (do not change, but be not the widest constant yardstick band, continuous S
i=0); The most stable (unchanged, but the widest yardstick that do not change, continuous S
i=0), so, we analyze the ration statistics of the best object yardstick of the SD variation that matches above-mentioned three types.Concrete design sketch is referring to Fig. 4.
Design sketch based in above-mentioned Fig. 4, the effect that optimal scale is extracted is analyzed accordingly.Concrete, what 8 class soils were covered cuts apart the statistics that middle best scale is selected, and truly mates yardstick and account for respectively 76%, 21% and 3% on the most stable, relatively stable, unstable yardstick.The true coupling yardstick majority of the water surface and broad-leaf forest is on the most stable yardstick, and they have homogeney in more class, than other soil, covers and shows obvious feature difference.And metastable SD reflects the otherness of the class inner structure in the multiple dimensioned variation that some soils cover conventionally.There is uncertainty the coupling aspect of the optimal scale of road object.Nearly half object is in the most stable commensurate in scope.Contiguous residence and arable land on the common space of road, and these types have similar spectral signature to road.These two reasons cause best true coupling yardstick not drop on all dropping on the most stable yardstick.But for most of soils, cover and area, utilize the most stable scale selection optimum matching object can obtain good segmentation effect.
In the present embodiment, by the true units match based on each object bounds, be that target is carried out cutting object extraction, realized the multi-scale division to image, the appearance of the overlap problem while having avoided the stack of multiple dimensioned object, has improved the accuracy of multi-scale division.
Through the above description of the embodiments, those skilled in the art can be well understood to the mode that the present invention can add essential general hardware platform by software and realize, and can certainly pass through hardware, but in a lot of situation, the former is better embodiment.Understanding based on such, the part that technical scheme of the present invention contributes to prior art in essence in other words can embody with the form of software product, this computer software product is stored in a storage medium, comprise that some instructions are with so that a computer equipment (can be personal computer, server, or the network equipment etc.) carry out the method described in each embodiment of the present invention.
It will be appreciated by those skilled in the art that accompanying drawing is the schematic diagram of a preferred embodiment, the flow process in accompanying drawing might not be that enforcement the present invention is necessary.
The invention described above embodiment sequence number, just to describing, does not represent the quality of embodiment.
Disclosed is above only a specific embodiment of the present invention, and still, the present invention is not limited thereto, and the changes that any person skilled in the art can think of all should fall into protection scope of the present invention.
Claims (6)
1. a multiple dimensioned stack dividing method, is characterized in that, said method comprising the steps of:
Image is carried out to multi-scale division, each yardstick object after being cut apart;
Utilization is stablized yardstick index and is judged each yardstick object, determines the best scale object of each yardstick;
The best scale object of each yardstick of determining is projected to single data Layer;
The best scale object projecting on single data Layer is merged.
2. the method for claim 1, is characterized in that, described image is carried out to multi-scale division, specifically comprises: based on region Fusion, from pixel, become object, small object to be fused into large object pixel fusion, merge step by step.
3. method as claimed in claim 2, is characterized in that, after each yardstick object that obtains cutting apart, in described each yardstick object, includes respectively characteristic parameter separately, and wherein, object standard deviation SD is the parameter of best scale identification.
4. method as claimed in claim 3, is characterized in that, described method also comprises:
What take over-segmentation object cuts apart yardstick at first as border, and co-located is cut the object of each yardstick, keeps each yardstick object characteristic parameter separately, so that different scale object carries out the contrast of co-located characteristic parameter.
5. method as claimed in claim 4, is characterized in that, the characteristic parameter of each yardstick object of yardstick index analysis is stablized in described utilization, specifically by following formula:
S
i＝F
i+1-F
i
Wherein, S
irefer to yardstick index of stability, i refers to level of zoom, F
ithe SD value of object on i yardstick, F
i+1to cut apart the object SD value in rank at i+1 yardstick;
In whole dimensional variation, S
ibe 0 continuously and continue when the longest, this interval medium scale is defined as best object fitting yardstick.
6. method as claimed in claim 5, is characterized in that, described the best scale object projecting on single data Layer is merged, and specifically comprises:
The characteristic parameter of single yardstick object of take is to merge according to the same value of carrying out adjacent object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410238491.3A CN104036499B (en) | 2014-05-30 | 2014-05-30 | Multi-scale superposition segmentation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410238491.3A CN104036499B (en) | 2014-05-30 | 2014-05-30 | Multi-scale superposition segmentation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104036499A true CN104036499A (en) | 2014-09-10 |
| CN104036499B CN104036499B (en) | 2017-02-22 |
Family
ID=51467258
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410238491.3A Expired - Fee Related CN104036499B (en) | 2014-05-30 | 2014-05-30 | Multi-scale superposition segmentation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104036499B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104881868A (en) * | 2015-05-14 | 2015-09-02 | 中国科学院遥感与数字地球研究所 | Method for extracting phytocoenosium spatial structure |
| CN105740825A (en) * | 2016-02-01 | 2016-07-06 | 福建师范大学 | Large-breadth remote sensing image building extraction method used for hybrid scene |
| CN106504259A (en) * | 2016-10-11 | 2017-03-15 | 昆明理工大学 | A kind of multiple dimensioned image partition method |
| CN107315914A (en) * | 2017-06-26 | 2017-11-03 | 太原理工大学 | A kind of multiple dimensioned nested spatial statistical units construction method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100322518A1 (en) * | 2009-06-23 | 2010-12-23 | Lakshman Prasad | Image segmentation by hierarchial agglomeration of polygons using ecological statistics |
| CN102609726A (en) * | 2012-02-24 | 2012-07-25 | 中国科学院东北地理与农业生态研究所 | Method for classifying remote sensing images blended with high-space high-temporal-resolution data by object oriented technology |
| CN103489193A (en) * | 2013-09-30 | 2014-01-01 | 河海大学 | High-resolution remote-sensing image change detection method facing targets and based on integrating strategy |
-
2014
- 2014-05-30 CN CN201410238491.3A patent/CN104036499B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100322518A1 (en) * | 2009-06-23 | 2010-12-23 | Lakshman Prasad | Image segmentation by hierarchial agglomeration of polygons using ecological statistics |
| CN102609726A (en) * | 2012-02-24 | 2012-07-25 | 中国科学院东北地理与农业生态研究所 | Method for classifying remote sensing images blended with high-space high-temporal-resolution data by object oriented technology |
| CN103489193A (en) * | 2013-09-30 | 2014-01-01 | 河海大学 | High-resolution remote-sensing image change detection method facing targets and based on integrating strategy |
Non-Patent Citations (2)
| Title |
|---|
| 林先成 等: "成都平原高分辨率遥感影像分割尺度研究", 《国土资源遥感》 * |
| 郭亚鸽: "面向对象的HJ-1CCD和TM影像土地覆盖信息提取研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104881868A (en) * | 2015-05-14 | 2015-09-02 | 中国科学院遥感与数字地球研究所 | Method for extracting phytocoenosium spatial structure |
| CN104881868B (en) * | 2015-05-14 | 2017-07-07 | 中国科学院遥感与数字地球研究所 | Phytobiocoenose space structure extracting method |
| CN105740825A (en) * | 2016-02-01 | 2016-07-06 | 福建师范大学 | Large-breadth remote sensing image building extraction method used for hybrid scene |
| CN105740825B (en) * | 2016-02-01 | 2018-11-23 | 福建师范大学 | It is a kind of for mixing the large format remote sensing image building extracting method of scene |
| CN106504259A (en) * | 2016-10-11 | 2017-03-15 | 昆明理工大学 | A kind of multiple dimensioned image partition method |
| CN106504259B (en) * | 2016-10-11 | 2019-02-05 | 昆明理工大学 | A kind of multiple dimensioned image partition method |
| CN107315914A (en) * | 2017-06-26 | 2017-11-03 | 太原理工大学 | A kind of multiple dimensioned nested spatial statistical units construction method |
| CN107315914B (en) * | 2017-06-26 | 2019-01-29 | 太原理工大学 | A kind of spatial statistical units construction method of multiple dimensioned nesting |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104036499B (en) | 2017-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106384081B (en) | Slope farmland extraction method and system based on high-resolution remote sensing image | |
| Ene et al. | Single tree detection in heterogeneous boreal forests using airborne laser scanning and area-based stem number estimates | |
| Zhou | An object-based approach for urban land cover classification: Integrating LiDAR height and intensity data | |
| Wang et al. | A random forest classifier based on pixel comparison features for urban LiDAR data | |
| Shendryk et al. | Bottom-up delineation of individual trees from full-waveform airborne laser scans in a structurally complex eucalypt forest | |
| AU2015324185B2 (en) | Multi-spectral image labeling with radiometric attribute vectors of image space representation components | |
| JP7157434B2 (en) | Forest resource information calculation method and forest resource information calculation device | |
| CN104881868A (en) | Method for extracting phytocoenosium spatial structure | |
| CN103093456A (en) | Corn ear character index computing method based on images | |
| US11170215B1 (en) | System and method for discriminating and demarcating targets of interest in a physical scene | |
| Pirotti et al. | A comparison of tree segmentation methods using very high density airborne laser scanner data | |
| Yang et al. | Delineating individual tree crowns in an uneven-aged, mixed broadleaf forest using multispectral watershed segmentation and multiscale fitting | |
| CN104036499A (en) | Multi-scale superposition segmentation method | |
| CN109766824B (en) | Active and passive remote sensing data fusion classification method based on fuzzy evidence theory | |
| Berveglieri et al. | Successional stages and their evolution in tropical forests using multi-temporal photogrammetric surface models and superpixels | |
| JP2015152340A (en) | Forest type analysis apparatus, forest type analysis method and program | |
| Bandyopadhyay et al. | Classification and extraction of trees and buildings from urban scenes using discrete return LiDAR and aerial color imagery | |
| CN107832849A (en) | The power line gallery 3-D information fetching method and device in a kind of knowledge based storehouse | |
| CN102073867A (en) | Sorting method and device for remote sensing images | |
| Jónsson | RGB and Multispectral UAV image classification of agricultural fields using a machine learning algorithm | |
| Wang et al. | Combining trunk detection with canopy segmentation to delineate single deciduous trees using airborne LiDAR data | |
| Mayr et al. | Mapping eroded areas on mountain grassland with terrestrial photogrammetry and object-based image analysis | |
| CN107103295A (en) | Optical remote sensing image cloud detection method of optic | |
| Kim et al. | Tree and building detection in dense urban environments using automated processing of IKONOS image and LiDAR data | |
| JP6334281B2 (en) | Forest phase analysis apparatus, forest phase analysis method and program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170222 Termination date: 20180530 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |